The Pentagon and Google have finalised an agreement enabling the US Defence Department to use the tech giant’s advanced Gemini AI systems within classified networks.
This official requested anonymity due to a lack of authorisation to reveal specifics regarding the agreement. The exact terms and conditions of the new contract remain unknown.
This agreement follows similar partnerships established with other prominent AI firms, such as OpenAI and xAI. Defence Secretary Pete Hegseth has put the integration of AI within the armed forces first, pledging to evolve the military into an “AI-first warfighting force.”
Over the last decade, the Defence Department has increasingly adopted AI technologies, employing automated systems for various tasks, including the analysis of drone footage in the campaign against the Islamic State group, improving logistics, and resolving pay discrepancies for service members.
Currently, AI is being used to assess intelligence and offer targeting assistance in the ongoing conflict with Iran.
Michael Horowitz, a former senior defense official and now a professor at the University of Pennsylvania, commented that the agreement to use Google’s AI models for classified applications highlights the escalating importance of AI in US national security.
Nevertheless, Horowitz pointed out that Google’s AI systems were already operational on unclassified platforms, making it unsurprising that an agreement was reached regarding classified applications.
In recent months, the Pentagon has been negotiating new contracts with the four largest AI companies in the United States, aiming to incorporate provisions that permit “any lawful use” of their AI technologies.
The Pentagon initially announced exploratory contracts with Google, OpenAI, Anthropic, and xAI in July.
These initiatives have generated some controversy, particularly concerning Anthropic.
The company, led by CEO Dario Amodei, has requested stronger assurances from the Pentagon to ensure that its AI models will not be employed for domestic mass surveillance or for the direct control of lethal autonomous weapons.
Read more: Google launches platform to build independent AI agents
